Constants and Variables 144, HRC vs non-HRC version
Various commentators are discussing the Iowa primary fiasco from a technical standpoint. Supposedly the app that was meant to replace paper and telephones failed because it hadn't been tested.

More likely it was intentionally built to create chaos and pave the way for Goddess Hillary to sweep in and settle the chaos, but we'll go with the story of poor testing for now.

This is the Silicon Valley norm. It's still not widely understood. It was described concisely by one of the Tesla skeptics here.

Share Value is Newthink. Testing products is Oldthink. Profit-seeking NEGATIVE EXTERNALITIES test their products because Oldthinkers want to have satisfied customers. Silicon Valley wants exponential Share Value, and the way to get there is by exponential growth of fake numbers. If you're tied down to products and employees and profits and other NEGATIVE EXTERNALITIES, you can't fake all the numbers.

As it happens, I'm in the middle of a nicely controlled experiment on the importance of testing.

My courseware comes in two versions, using the same text and animations but somewhat different JS/HTML active code. One version is designed to run inside e-learning LMS systems, and the other is a non-graded study guide. The differences are minor, but enough to require handling them separately in my own processing.

The publisher has been carefully proofreading and testing the LMS versions because grades have consequences. The non-LMS version wasn't carefully tested. Though the products have been out for a couple of semesters, this semester seems to be the first time instructors are really adopting and trying them, and some errors are appearing in the non-LMS version that wasn't thoroughly tested beforehand.

I'm hustling to handle the errors, but not hustling TOO fast. The current publisher is a small profit-making business, not a giant Share Value conglomerate. I respect them and want them to succeed. Debugging too fast generally introduces NEW bugs. I'd rather have a little extra delay.

= = = = =

A bit later.... This turned out to be a hugely important blog item for my purposes. It opened up a big question. Do I really NEED to maintain two versions? The code in the LMS version works better for the specific error reported yesterday. Instead of trying to untangle where the non-LMS fails, can I just interrogate the surroundings to see if an LMS is present, and then use the resulting variable to switch in ONLY the pieces that actually call the LMS? The answer is yes.

Labels: Constants and Variables

  ¶ 11:57 PM